fix

huggingface · Apr 18, 2024 · b58676f · b58676f
1 parent 138ff00
commit b58676f
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 0 deletions.
diff --git a/src/transformers/models/clipseg/modeling_clipseg.py b/src/transformers/models/clipseg/modeling_clipseg.py
@@ -736,6 +736,7 @@ def forward(
             pooled_output = last_hidden_state[
                 torch.arange(last_hidden_state.shape[0], device=last_hidden_state.device),
                 # We need to get the first position of `eos_token_id` value (`pad_token_ids` might equal to `eos_token_id`)
+                # Note: we assume the input always has a eos token in each text (i.e. always prepared by clipseg tokenizer)
                 (input_ids.to(dtype=torch.int, device=last_hidden_state.device) == self.eos_token_id)
                 .int()
                 .argmax(dim=-1),

diff --git a/src/transformers/models/groupvit/modeling_groupvit.py b/src/transformers/models/groupvit/modeling_groupvit.py
@@ -1118,6 +1118,7 @@ def forward(
             pooled_output = last_hidden_state[
                 torch.arange(last_hidden_state.shape[0], device=last_hidden_state.device),
                 # We need to get the first position of `eos_token_id` value (`pad_token_ids` might equal to `eos_token_id`)
+                # Note: we assume the input always has a eos token in each text (i.e. always prepared by clip tokenizer)
                 (input_ids.to(dtype=torch.int, device=last_hidden_state.device) == self.eos_token_id)
                 .int()
                 .argmax(dim=-1),